AI's Next Leap: Why Independent Oversight Can't Wait

Jan de Vries · 2026-10-09

Vendan Ananda Kumararajah recently dropped a wake-up call that's worth pausing over. He argues that AI governance needs independent oversight—not just to keep pace with the technology, but to stop flawed assumptions from multiplying as increasingly capable systems start building their own successors. In plain terms: if we don't get this right, we risk an intelligence explosion we can't control. That's not a sci-fi plot. It's a governance gap. And it's widening fast. ### The Problem with Self-Regulation Right now, many AI labs operate like kids running a candy store. They set their own rules, decide what's safe, and grade their own homework. Kumararajah points out that this approach is dangerously circular. When the same people who build the systems also define the safety standards, blind spots become baked in. And those blind spots don't stay small—they compound. Imagine a self-driving car that only tests itself on sunny days. It might feel safe until the first rainstorm. Now scale that to an AI that can rewrite its own code. The stakes aren't just a fender bender; they're existential. ### Why Independent Oversight Isn't Optional Independent oversight means bringing in outside experts—ethicists, engineers, policymakers—who have no financial stake in the AI's success. Their job? To challenge assumptions, audit decisions, and flag risks before they become catastrophes. It's like having a referee in a boxing match. Without one, the fighters might not even notice they're breaking the rules. Kumararajah isn't calling for a bureaucratic nightmare. He's asking for a simple check-and-balance system. One that evolves as fast as the tech does. > "The real danger isn't that AI becomes evil," Kumararajah suggests. "It's that we become lazy—assuming someone else is watching." ### What Happens If We Ignore This? Here's the uncomfortable truth: if we don't build independent oversight now, we'll be playing catch-up later. And catch-up in AI isn't like catching up on laundry. It's more like trying to stop a runaway train with a butterfly net. - **Flawed assumptions multiply.** A small error in an AI's logic can replicate across millions of decisions in seconds. - **Accountability vanishes.** When no one's in charge, everyone points fingers. - **Public trust erodes.** People already worry about AI. A major scandal could set the field back decades. We've seen this movie before—with social media, with crypto, with data privacy. Each time, we waited until the damage was done. Let's not do that with something that could outthink us. ### A Better Way Forward So what does good governance look like? Kumararajah offers a few starting points: - **Independent review boards** with real power, not just advisory roles. - **Transparency mandates** so outsiders can see how decisions are made. - **Global coordination** because AI doesn't respect borders. - **Adaptive rules** that update as the technology evolves. None of this is easy. But easy isn't the point. The point is survival—of our values, our safety, and our say in the future. Look, I get it. Governance sounds boring. It's not as flashy as a new model release. But boring is what keeps the lights on. Boring is what stops the explosion. And right now, boring might be the most important thing we do. Kumararajah's message is clear: rethink AI governance, or risk an intelligence explosion we can't undo. The choice is ours. Let's not blow it.